As we discussed before, with Iverson on the floor Sixers won "a lot" more (as pointed out by Mike G. +19 wins per season average in his 1996-2006 sixers career which is truly astronomic) and gained a significant boost in TS% and decrease in TO%.
I also mentioned, Sixers won more when Iverson shot more. For details I referred to this post from 2005.
http://www.basketballforum.com/nba-stat ... empts.html
Randomly, I came upon an another work of someone else from 2011 that analyses 9 players' primes (between 4 and 9 years depending on the player) and eventually prove Sixers won more when Iverson shot more. But what's more interesting is out of 9 great wing scorers in NBA history (Iverson, Jordan, Kobe, T-mac, Carter, LeBron, Wade, Arenas, Wilkins), Iverson's the only one whose team got better when he shot more although his TS% is the worst out of those 9 and the second worst in games with 30+ fga. On the contrary Wade and Kobe hurt their teams really bad when they shot more.
https://elgee35.wordpress.com/2011/02/1 ... -part-iii/
With all of these I think Iverson is a big outlier for today's metrics that every analyst should be aware of and take lessons from it since there can be player actions that impact the game considerably but even the new player-tracking cameras can't capture. Advanced metrics are not at that point that beats "biased" eye test of people that know and watch the game, yet. There's a reason with 100% roster turnover, PER (which is a really bad metric that I never used it in my blends) beats xRAPM.
http://ascreamingcomesacrossthecourt.bl ... trics.html
BTW a little bit out of topic but this made me wonder why do people here use xRAPM, RPM, BPM or WS to evaluate player greatness when PER (which I don't like) is better than them at out-of-sample prediction accuracy with 100% roster turnover? I hate PER but if it's more accurate than other metrics at out-of-sample prediction with 100% roster turnover, it means it's better when it comes to value players solely regardless of the teams, teammates and the system they play in. Other metrics I mentioned will beat PER always at predicting at team level because of low roster-turnover rate but it's a whole different story here.